[Back]

A. Ademaj, A. Hanzlik, H. Kopetz:
"Tolerating Arbitrary Failures in a Master-Slave Clock-Rate Correction Mechanism for Time-Triggered Fault-Tolerant Distributed Systems with Atomic Broadcast";
Talk: International Conference on Real-Time and Network Systems (RTNS), Nancy, Frankreich; 2007-03-29 - 2007-03-30; in: "Proceedings of the 15th International Conference on Real-Time and Network Systems (RTNS'07)", Institut National Polytechnique de Lorraine, Nancy, Frankreich (2007), ISBN: 2-905267-53-4; 215 - 224.

In a previous work we have shown that by deploying a
node with a high-quality oscillator (rate-master node) in
each cluster of a real-time system, we can integrate internal
and external clock synchronization by a combination
of a distributed mechanism for clock state correction with
a master-slave mechanism for clock rate correction. By
means of hardware and simulation experiments we have
shown that this combination improves the precision of the
global time base in single- and multi-cluster systems while
reducing the need for high-quality oscillators for non ratemaster
nodes. Previous experimental results have shown
that transient fail-silent failures of the rate-master node
will not affect the operation of the distributed clock state
correction algorithm, and the cluster will remain internally
synchronized. In this paper we consider all possible
failure modes of the rate-master node by taking the system
structure into account, and present a solution that tolerates
arbitrary rate-master failures by using replicated
rate-master nodes. Possible arbitrary failure modes are
listed in the paper where a main part of them is handled
by the fault-tolerant mechanisms of the system architecture,
whereas the remaining failure modes are handled by
a fault-tolerant rate correction mechanism.